Frequency of extreme freeze events controls the distribution and structure of black mangroves (Avicennia germinans) near their northern range limit in coastal Louisiana

Climate change is expected to result in the tropicalization of coastal wetlands in the northern Gulf of Mexico, as warming winters allow tropical mangrove forests to expand their distribution poleward at the expense of temperate salt marshes. Data limitations near mangrove range limits have hindered understanding of the effects of winter temperature extremes on mangrove distribution and structure. Here, we investigated the influence of extreme freeze events on the abundance, height and coverage of black mangroves (Avicennia germinans) near their northern range limit in Louisiana.


| INTRODUC TI ON
In the face of rapid climate change, ecologists and natural resource managers are increasingly challenged to better anticipate and prepare for the effects of changing temperature and precipitation regimes on the distribution of species, ecosystems and biomes (Pecl et al., 2017;Scheffers et al., 2016;Thomas, 2010). Near the transition between tropical and temperate climates, warming air and ocean temperatures are expected to allow tropical biomes to move poleward-towards the north pole in the northern hemisphere and towards the south pole in the southern hemisphere (Carter et al., 2018;Chen, Hill, Ohlemüller, Roy, & Thomas, 2011;Day et al., 2013;Parmesan, 2006;Yamano, Sugihara, & Nomura, 2011).
Scientists have used the term "tropicalization" to describe the transformation of temperate ecosystems by these poleward-moving tropical species (Macy et al., 2019;Scheffel, Heck, & Johnson, 2018;Vergés et al., 2014;Yáñez-Arancibia, Day, Twilley, & Day, 2014). In coastal wetland ecosystems, warming winter air temperatures are expected to allow tropical mangrove forests to move poleward, in some cases at the expense of temperate salt marsh ecosystems (Cavanaugh et al., 2019;Saintilan, Wilson, Rogers, Rajkaran, & Krauss, 2014). The ecological implications of these marsh-to-mangrove transformations are large (Guo et al., 2017;Kelleway et al., 2017), and ecologists working near mangrove range limits are increasingly challenged to better understand the drivers and implications of mangrove range expansion. Here, we investigated the influence of winter air temperature regimes on the distribution and structure of mangroves near a northern range limit in wetland-rich Louisiana (USA).
However, the influence of winter temperature regimes varies greatly within and across range limits (Cavanaugh et al., 2018;Cook-Patton, Lehmann, & Parker, 2015;Osland, Feher, et al., 2017;Quisthoudt et al., 2012). While some range limits (e.g., eastern North America) are controlled by extreme freeze events (e.g., individual freeze events that may occur for just a few days once every decade or two), others are controlled by consistently cold temperatures that are slightly above freezing for the entire winter (e.g., Australia, South Africa, South America, New Zealand; Cavanaugh et al., 2018;Morrisey et al., 2010;Osland, Feher, et al., 2017;Stuart et al., 2007).
Coastal wetland scientists working in eastern North America have long recognized that the frequency of extreme freeze events governs the northern range limits of mangrove forests in Texas, Louisiana and Florida (Kennedy, Preziosi, Rowntree, & Feller, 2020;Lloyd & Tracy, 1901;Lonard & Judd, 1991;Lugo & Patterson-Zucca, 1977;Sherrod & McMillan, 1985;West, 1977). However, extreme freeze events that lead to mangrove mortality or damage are infrequent and difficult to study (Osland, Day, et al., 2020;Pickens, Sloey, & Hester, 2019;Ross et al., 2009). Moreover, mangrove distribution data have historically been lacking near northern range limits in this region (Armitage, Highfield, Brody, & Louchouarn, 2015;Giri & Long, 2014Osland et al., 2018). Thus, the influence of extreme freeze events on mangrove distribution and structure has been poorly quantified.
There are three common mangrove species in North America: Avicennia germinans, Laguncularia racemosa and Rhizophora mangle (Tomlinson, 1986). Of these three species, A. germinans (black mangrove) is the species that is most freeze-tolerant and the species whose distribution extends farthest north in eastern North America.
To our knowledge, A. germinans is the only mangrove species that is currently present in Louisiana. In this study, we investigated the following questions for the northern range limit of A. germinans in Louisiana: (1) Within the past 30 years (i.e., 1989-2018), what has been the frequency and spatial distribution of extreme freeze events with the potential to cause A. germinans mortality and/or leaf damage? (2) Where is A. germinans located and how is its distribution influenced by the spatial distribution of extreme freeze events?
(3) What are the relationships between the frequency of extreme freeze events and the abundance, height and coverage of A. germinans? and (4) How does the risk of A. germinans freeze damage vary spatially across coastal Louisiana? Advancing knowledge on these topics will help scientists and natural resource managers better anticipate and prepare for the tropicalization of coastal wetlands due to climate change.

| Study area background: coastal Louisiana
Coastal Louisiana is positioned within the dynamic Mississippi River Delta, which is one of the largest river deltas in the world. In addition to providing valuable fish and wildlife habitat, Louisiana's abundant wetlands offer protection from storms, improve water quality, sequester carbon from the atmosphere, support food K E Y W O R D S Avicennia germinans, black mangrove, climate change, coastal wetland, freezing, Louisiana, mangrove, range limit, salt marsh, temperature webs, provide seafood, store floodwaters and provide recreational opportunities (Barbier et al., 2011;Costanza et al., 2014).
Though difficult to quantify, the societal benefits (i.e., ecosystem services) provided by coastal ecosystems in Louisiana's Mississippi River Delta have been valued to be at least $12-$47 billion (US dollars) per year (Batker et al., 2010). However, the rate of wetland loss in Louisiana has been very high in the past century, due to a combination of natural and human factors that have reduced the ability of wetlands to build elevation to keep pace with high rates of subsidence and relative sea-level rise (Blum & Roberts, 2009;Day et al., 2007;Törnqvist, Jankowski, Li, & González, 2020).
Between 1932 and 2016, Louisiana lost approximately 4,833 km 2 of wetlands (Couvillion, Beck, Schoolmaster, & Fischer, 2017), and the state has become a prominent global example of the negative linkages between high relative sea-level rise, sediment delivery alterations and coastal wetland loss (Jankowski, Törnqvist, & Fernandes, 2017;Twilley et al., 2016). Given the rapid pace of wetland loss in the last century and the expectation of accelerated  (Carter et al., 2018;USGCRP, 2017USGCRP, , 2018. Winter warming is a critical component of climate change that is expected to transform Louisiana's coastal ecosystems (Gabler et al., 2017;Osland, Enwright, Day, & Doyle, 2013). Extreme freeze events have historically served as major ecological disturbances in Louisiana and other tropical-temperate transition zones across North America-leading to mass mortality of freeze-sensitive organisms. Freeze events control the distribution of mangrove forests in Louisiana  and across the northern Gulf of Mexico and Atlantic coasts of North America (Cavanaugh et al., 2014;Osland et al., 2013;Ross et al., 2009;Sherrod & McMillan, 1985;Stevens, Fox, & Montague, 2006). Most of Louisiana's salt-affected tidal wetlands (i.e., those in the saline, brackish and intermediate salinity classes) are currently dominated by freeze-tolerant salt marsh graminoid plants (i.e., grasses, sedges and rushes; Osland, Grace, et al., 2019;Sasser, Visser, Mouton, Linscombe, & Hartley, 2014;Visser, Sasser, Chabreck, & Linscombe, 1998, 2000. Due to their current and historical role as critical foundation plant species (sensu Ellison, 2019; Ellison et al., 2005), these graminoid-dominated salt marsh plant communities tend to be the target of future-focused ecological assessments and restoration-focused planning efforts for coastal Louisiana However, the outer coast of Louisiana contains a dynamic and climate-sensitive mangrove-marsh ecotone that has the potential to be transformed by warming winters (Gabler et al., 2017;Osland et al., 2013).
How long have mangroves been present in Louisiana? The recent ~30-year expansion of mangroves since the last major, mortality-inducing freeze event in 1989 happens to coincide with the availability of remotely sensed Landsat data; thus, there has been some debate regarding the novelty of the last 30 years of mangrove expansion in eastern North America (Armitage et al., 2015;Cavanaugh et al., 2014Cavanaugh et al., , 2019Giri & Long, 2014Saintilan et al., 2014) (Romans, 1775). In 1779, French surveyor Jean Francois Gonsoulin described mangroves near Cat Island Pass during an effort to find a coastal navigation route between New Iberia and New Orleans (Weddle, 1995 Bent, 1910;Brown, 1930;Kopman, 1915;Moore, 1899;Penfound & Hathaway, 1938), and digital herbarium records of Avicennia in Louisiana, available on the Southeast Regional Network of Expertise  Bartram (Bartram & Harper, 1943) and André Michaux (Michaux & Sargent, 1889), respectively.
An analysis that used historical temperature records dating back to 1890s and mangrove abundance data near Port Fourchon dating back to the 1970s indicates that the mangrove range limit in Louisiana has expanded and contracted many times in response to the absence or presence of winter temperature extremes, respectively . During warm periods, freeze-sensitive mangroves have expanded as they outcompete salt marsh graminoids. Conversely, following extreme freeze events that kill or damage mangroves, the coverage of freeze-tolerant salt marshes has expanded. Similar mangrove expansion and contraction cycles have also occurred in Florida (Cavanaugh et al., 2019;Stevens et al., 2006) and other mangrove range limits across eastern North America (Kennedy et al., 2016(Kennedy et al., , 2017Sherrod & McMillan, 1981, 1985. Since the last regionally relevant major freeze event occurred in December 1989, the mangrove-marsh ecotone has been expanding northward in Louisiana and across the region (e.g., in coastal Texas and both coasts of the Florida peninsula). Mangrove range limits in the region are expected to continue to expand farther north in response to warming winters . In the face of climate change, mangroves are increasingly seen as potential vegetation targets for coastal wetland restoration and planning efforts (Hijuelos et al., 2019;Mack et al., 2014;USDA-NRCS, 2017). For example, mangrove seedlings are sometimes planted at restoration sites, and mangrove propagules have also been occasionally collected and dispersed into salt marshes. Mangroves may also be a target vegetation that would establish within planted marsh vegetation following natural recruitment of mangrove propagules from adjacent mangrove stands. However, to better inform restoration practices, there is a need to better understand the influence of extreme freeze events upon mangrove distribution and structure near this range limit.

| Study grid creation
Our study area included all of coastal Louisiana (Figure 1), which can be coarsely divided into two geomorphic regions: the Mississippi River Deltaic Plain (Bahr et al., 1983) to the east and the Chenier

| Mangrove data
We incorporated A. germinans abundance, height and cover data for the year 2009 from a U.S. Geological Survey Data Release (Day, Michot, Twilley, & From, 2020

| Station-based temperature data
Our final freeze frequency analyses relied on gridded temperature data; however, we used station-based daily minimum temperature data to identify freeze events for which gridded data were obtained and analysed in more detail.  . Thus, using this temperature cut-off for an inland location (i.e., Lafayette) provided us with a conservative dataset for quantifying the frequency of events with the potential to lead to mangrove mortality or damage along the slightly warmer coast, where temperatures during these events should fall below approximately −3.8°C. Avicennia germinans leaf damage in this area typically begins at temperature less than −4.2°C (Osland, Day, et al., 2020).

| Gridded temperature data
For each of the 19 days that were identified as having potentially ecologically relevant minimum temperatures, we obtained continuous gridded daily minimum air temperature data that were created by the PRISM Climate Group using the PRISM (Parameter-elevation Relationship on Independent Slopes Model) interpolation method (Daly et al., 2008). The PRISM data were selected because the PRISM model accounts for land-ocean temperature gradients (Daly et al., 2008;Daly, Helmer, & Quiñones, 2003;Daly, Widrlechner, Halbleib, Smith, & Gibson, 2012) that influence spatial patterns of mangrove mortality and damage ).

| Data analyses: freeze frequency
We defined potential mangrove mortality and leaf damage events as freeze events with temperatures below −6.6°C and between F I G U R E 1 Map of Avicennia germinans (black mangrove) distribution in Louisiana (USA). Each of the orange circles represents a mangrove observation point recorded via aerial surveys from a fixedwing aircraft in 2009 Michot et al., 2010). The seven coastal vegetation types are from Sasser et al. (2014). Winter air temperature extremes (i.e., freeze events) constrain Louisiana mangroves primarily to saline and brackish marshes of the south-eastern outer coast −4.2°C and −6.6°C, respectively (Osland, Day, et al., 2020). For each cell within the study grid and for each mangrove observation point, we used the daily gridded climate data to determine the number of events (i.e., the frequency of extreme freeze events) that would lead to mangrove mortality or leaf damage during 30-year (1989-2018), 20-year (1999-2018) and 10-year periods (2009-2018).

| Data analyses: freeze-mangrove relationships and spatial depictions of mangrove risk
We used regression analyses to quantify the relationships between 10-year periods (2009)(2010)(2011)(2012)(2013)(2014)(2015)(2016)(2017)(2018). Finally, we produced maps depicting the risk of mangrove freeze damage using the 30-year frequency of mangrove damage events. Moderate, high and very high categories for risk of mangrove freeze damage were assigned to areas that were determined to have 2 to 3, 4 to 5, and ≥6 potential leaf damage events during the 30-year period, respectively. All spatial analyses were conducted using ArcGIS 10.6 (Environmental Systems Research Institute, Inc.). All regression analyses were conducted in SigmaPlot 12.5 (Systat software, Inc.).

| Severe freeze events between 1989 and 2018
For the 30-year period between 1989 and 2018, our analyses identi- Potential mortality and leaf damage events were defined as freeze events with temperatures below −6.6°C and between −4.2°C and −6.6°C, respectively (Osland, Day, et al., 2020). Note that these maps represent events with the potential to affect mangroves regardless of whether they are present. See Figure 1 for current mangrove distribution 2010 and 2018 events, the temperatures were much less severe than the 1989 Christmas freeze but cold enough to result in some mangrove mortality and leaf damage in certain areas (note red and yellow areas in Figure 2b-

| Frequency of extreme freeze events
We used the gridded temperature data to produce maps of the frequency of freeze events with the potential to cause mangrove mortality and mangrove leaf damage in Louisiana across 30-year ( Figure 3a,b), 20-year (Figure 3c,d) and 10-year periods (Figure 3e,f).
These maps corroborate the hypothesis that temperatures are colder in the Chenier Plain (compared to the Deltaic Plain) and in inland Louisiana (compared to coastal Louisiana). Note that the mangrove observation points shown in Figure 1 are in areas where the frequency of extreme freeze events is lowest (i.e., the south-eastern outer coast).

| Mangrove abundance, height, and cover and the influence of freeze frequency
The aerial surveys identified 4,382 mangrove observation points, which were mostly located along the south-eastern outer coast in areas that are classified as saline or brackish marshes in Sasser et al.
(2014) (Figure 1). The potential for freeze-induced mangrove mortality is higher along the Chenier Plain compared to the outer coast of the Deltaic Plain (Figures 2 and 3).

F I G U R E 3
The frequency of freeze events with the potential to cause mangrove mortality (left panels) and mangrove leaf damage (right panels) in Louisiana. Frequency was quantified across 30-year (a, b), 20-year (c, d) and 10-year periods (e, f). These three periods include freeze events that occurred between 1989 and 2018, 1999 and 2018, and 2009 and 2018, respectively. Potential mortality and leaf damage events were defined as freeze events with temperatures below −6.6°C and between −4.2°C and −6.6°C, respectively (Osland, Day, et al., 2020). Note that these maps represent events with the potential to affect mangroves regardless of whether they are present. See Figure 1 for mangrove distribution in 2009 height categories, respectively; Day et al., 2020;Michot et al., 2010).
The potential for freeze-induced mangrove mortality is higher in inland areas compared to the outer coast (Figures 2 and 3).   (Osland et al., 2015;Pickens et al., 2019).

| D ISCUSS I ON
For the A. germinans range limit in Louisiana, our findings quantify the influence of the frequency of extreme freeze events upon

F I G U R E 4
The relationships between the frequency of freeze events and the number of mangrove observation points in the following: (a, b) all mangrove categories; (c, d) the tall mangrove category (i.e., height greater than 2 m); and (e, f) the solid mangrove cover category (i.e., groups of several plants up to continuous mangrove cover in areas up to 7,850 m 2 ). The mangrove observation points were recorded via aerial surveys from a fixed-wing aircraft in 2009 Michot et al., 2010). The x axes represent the number of freeze events with the potential to cause mangrove mortality (left panels) or leaf damage ( Texas (Sherrod & McMillan, 1981, 1985 and Florida (Davis, 1940;Olmsted, Dunevitz, & Platt, 1993;Stevens et al., 2006). Literature observations (Table 1), herbarium records (Table 1), historical landscape-level plant surveys (e.g., Chabreck, 1970;Visser et al., 1998;Visser et al., 2000),  In addition to the risk that planted seedlings will be killed by extreme freezing temperatures, there is increasing evidence that salt marsh grasses (e.g., Spartina alterniflora) are able to more effectively jump-start coastal wetland restoration efforts compared to mangroves due to the more rapid growth, horizontal expansion and recruitment of marsh grasses compared to mangroves (specifically, A. germinans; Yando, 2018;Yando, Osland, Jones, & Hester, 2019). Once established, planted salt marsh grasses can help trap mangrove propagules and facilitate the natural recruitment and growth of A. germinans seedlings (Donnelly, Walters, & coasts, 2014;Osland, Feher, Spivak, et al., 2020;Peterson & Bell, 2012

| Climate change implications
Our map of mangrove observations points (Figure 1)  will serve as hotspots for future mangrove expansion.
Since the last major freeze event occurred in 1989, A. germinans individuals in these areas have had thirty years to grow, expand, and become more resistant and resilient to future freeze events. Due to positive feedbacks between vegetation height and temperature, larger mangroves are typically more resistant to freezing temperatures than smaller mangroves (D'Odorico et al., 2010(D'Odorico et al., , 2013Osland et al., 2015;Osland, Hartmann, et al., 2019;Weaver & Armitage, 2018). Larger mangroves also produce more propagules (Alleman & Hester, 2011), which can enhance the ability of mangrove populations to recover from freeze events via natural regeneration.
During extreme freeze events, there can be dramatic vertical mangrove damage gradients due to temperature differences between air, water and soil; for example, air temperatures near the soil surface can be ~5°C warmer than at 25 cm above the soil surface (Osland, Hartmann, et al., 2019). As a result, mangrove propagules lying on the soil surface are often protected from freeze events by the buffering effects of warmer soil temperatures. So, following freeze events that kill more exposed mangrove seedlings and trees, mangrove natural regeneration can occur from the propagules lying on the soil surface. In response to warming winters in the coming century, we expect that mangroves will expand into Louisiana's interior Deltaic Plain and Chenier Plain; however, we do not expect that mangrove expansion will occur in a steady unidirectional manner. Rather, we expect that there will be pulses of rapid range expansion during freeze-free years followed by abrupt periods of range contraction after extreme freeze events. Due to the presence of vegetationmicroclimate feedbacks, the rate of mangrove range expansion has the potential to accelerate during extended freeze-free periods (Osland et al., 2015). To complicate matters, mangrove range expansion will also be influenced by interactions with other aspects of climate change (McKee, Rogers, & Saintilan, 2012;Osland et al., 2018;Saintilan, Rogers, & McKee, 2019;Ward et al., 2016).

| Knowledge gaps, research needs and data limitations
In this study, we used the best available data to show that spatial variation in freeze frequency governs mangrove abundance, height and coverage in Louisiana. Our analyses provide a foundation for advancing understanding of climatic controls on the distribution of mangroves in one of the largest wetland complexes in the world.
However, our work also shows that there is a need for improved  (Alleman & Hester, 2011;Yando, 2018). Models of future mangrove expansion will require a better understanding of mangrove propagule dispersal across the Louisiana coast.
Although extreme freeze events across land-ocean temperature gradients play a critical role in restricting the distribution of mangroves in Louisiana, there are other concomitantly varying biotic and abiotic factors along these gradients that will also affect the future distribution and structure of mangroves. Within the context of mangrove expansion, these factors warrant additional consideration.
For example, there are inundation, salinity and surface elevation change gradients across this zone that affect plant-plant interactions (Howard et al., 2015(Howard et al., , 2020Jiang et al., 2016). There are also concomitant gradients in mangrove propagule density, dispersal limitation and perhaps herbivory pressure, which may hinder rapid mangrove migration into interior areas that become climate-suitable. In less saline and more highly inundated wetlands that contain tall marsh vegetation, mangroves may not immediately be able to become established and compete for light and other resources.
Mangrove establishment within these highly productive marshes may require ecological disturbances (e.g., drought, fire, prolonged inundation, hypersalinity, hurricanes, extreme grazing pressure) that lead to marsh vegetation dieback, producing space for mangrove dis-

| CONCLUSIONS
Our analyses advance understanding of how the frequency of extreme freeze events controls the distribution, height and coverage of A. germinans near its northern range limit in wetland-rich Louisiana. Avicennia germinans is most abundant, tall and continuous along the south-eastern outer coast of Louisiana, where there is a lower risk of mangrove freeze damage (i.e., lower frequency of extreme freeze events). In contrast, the risk of A. germinans freeze damage has historically been very high across Louisiana's Chenier Plain and within more inland wetlands in the Deltaic Plain. In addition to informing climate-smart coastal restoration efforts, our findings can be used to better anticipate and prepare for the tropicalization of temperate wetlands due to climate change.

DATA AVA I L A B I L I T Y S TAT E M E N T
The data used in this study are all publicly available. The mangrove observation points are available in Day et al. (2020